Cloud Bat Intelligence

What is Yunbat Intelligence?

Cloud Bat Intelligence is a suite of large-model voice intelligents and AICC customer contact centers designed for enterprises.

It handles outbound calls, inbound calls, customer service, follow-ups, lead tracking, and call quality assessment by leveraging large models, RAG, and voice technology.

Main functions

• Create AI outbound call tasks in bulk.
• Use large models to process multi-turn phone conversations.
• Provides intelligent call-in services for incoming calls.
• Route to a human agent in complex scenarios.
• View call status and statistics in real time.
• Automatically identify customer intentions and tags.
• Perform redialing and subsequent actions according to the rules.
• The results are sent via WeChat or text message.
• Use large models to verify call quality.
• Manage leads, open ocean opportunities, and sales follow-ups.
• Free CRM and team collaboration tools are provided.
• Supports API, OEM, and private deployment.

Product matrix

Which companies are suitable?

• Retail businesses that need to conduct bulk customer follow-ups.
• A team responsible for handling logistics notifications and after-sales services.
• The brand management team responsible for recalling members.
• A call center that requires unified customer service standards.
• Public service agencies responsible for handling policy notifications.
• An educational institution that manages admissions counseling.
• Financial and collection teams that require call quality assessment.
• Software providers that wish to incorporate voice capabilities.

How does AI-powered outbound calling work?

After the company enters legally obtained numbers and sets up scripts, a knowledge base, and call times, the system makes calls automatically based on the assigned tasks.

During the call, it identifies the user’s statements, selects appropriate replies, assigns tags, and routes the call to a human agent or carries out further actions according to predefined rules.

1. Clarify the legitimate business purpose of this outbound call.
2. Organize the list of authorized customers.
3. Design the main workflow and branches for common issues.
4. Configure the knowledge base, intent tags, and variables.
5. Select ASR, TTS, and an appropriate voice tone.
6. Set call times and rules for redialing.
7. Use a small number of internal numbers to conduct the test.
8. Check for interruptions, delays, and recognition accuracy.
9. Observe task data in real time after it goes online.
10. Transfer complex or high-intent calls to a human agent.

Design of scripts for large models

Large-model scripts are more flexible than fixed nodes, but it is necessary to define business boundaries and the scope of knowledge that can be addressed.

Knowledge base and RAG

The knowledge base provides vetted business information for voice agents, while RAG is used to retrieve relevant content during calls.

• Only verified and valid materials should be uploaded.
• Broken down by product, region, and business type.
• Indicate the effective dates for prices and policies.
• Remove conflicting, duplicate, and expired data.
• Set up manual transfer for sensitive issues.
• Use real-world problems to continuously test the recall results.
• After each update, re-check the call responses.

Human-robot collaboration

Human agents can monitor multiple AI-powered calls simultaneously and take over proactively in high-value, complaint-related, or complex scenarios.

1. Set clear conditions for triggering a handover to a human agent.
2. Arrange a sufficient number of online human agents.
3. Synchronize customer and call information before transferring.
4. The agent can view the communication records that have been completed by AI.
5. Continue to address genuine needs after taking over.
6. Record the results and refine the customer tags.
7. Analyze the reasons why the transfer was not successful.

Smart incoming calls

Smart inbound calls are used to automatically answer customer calls, handling common inquiries, creating service tickets, and categorizing requests.

Call history and data statistics

The platform allows filtering of records by task, time, number, region, tag, duration, and number of call rounds.

• View call volume, connection rate, and call duration.
• Statistically analyze the distribution of potential customers and tags.
• Analyze the locations where node hits occur and where connections are terminated.
• View call peaks and customer locations.
• Record the results of transferring to manual handling and WeChat notifications.
• Export call records and task statistics.
• Calculate the cost of acquiring customers based on billing data.

Recall and blacklist

The system allows the setting of rules for redialing in cases such as rejected calls, device shutdown, machine shutdown, and line interception.

When a customer explicitly refuses, complains, or requests that further contact be ceased, they should be added to the blacklist immediately, and no further marketing attempts should be made toward them.

• Limited retries are set for technical disconnections.
• Avoid making repeated calls in a short period of time.
• Set compliance periods by region and business.
• Record the time when the customer refused to be contacted.
• Synchronize the blacklist across different tasks.
• Regularly audit the rules for releasing items from the blacklist.

Large-model voice quality inspection

Voice quality inspection can convert recordings into text and analyze them based on criteria such as keywords, semantics, emotion, silence, and interruptions.

Audio supported by quality inspection

The official help page lists formats such as WAV, MP3, M4A, WMA, AAC, OGG, AMR, and FLAC.

• A single file should be kept under 512MB.
• Supports single-track and dual-track recording.
• The processing time for long recordings will increase accordingly.
• The identification results should be manually corrected via sampling.
• Before uploading, verify that you have the permissions to process the recording.
• Sensitive recordings are encrypted and stored in accordance with the company’s policies.

CRM and lead follow-up

A free CRM for managing leads, prospects, responsible persons, tags, follow-up plans, contracts, and the sales funnel.

1. Import customer leads through legitimate channels.
2. Group by region, business, and intent.
3. Automatically write the results of outbound calls into the customer records.
4. Assign leads with high intent to sales representatives.
5. Set the plan for the next follow-up.
6. Record text messages, manual calls, and the results of visits.
7. Regularly clean up invalid and expired leads.

Complete tutorial for Yunbat Intelligence

1. Register an account and request a product demonstration.
2. Confirm the business scenario, call volume, and completion cycle.
3. Verify the circuit, qualifications, and compliance requirements for data.
4. Prepare the legally obtained numbers and customer fields.
5. Create script and business knowledge bases.
6. Configure the timbre, recognition engine, and dialogue parameters.
7. Set up intent tags and conditions for switching to a human agent.
8. Set call times, retry options, and block lists.
9. Conduct multiple rounds of testing using internal numbers.
10. Confirm the recording prompt and the disclosure of the AI’s identity.
11. Start real tasks in small batches.
12. Observe the connection rate, complaints, and intended outcomes.
13. Listen to the recordings and adjust the wording.
14. Transfer valid leads to CRM for follow-up.
15. Generate compliance audit and expense reports.

Price and cost breakdown

Cloud Bat Intelligence primarily provides enterprise-based pricing; the final cost depends on the number of calls, the lines used, the scripts employed, the method of deployment, and any custom requirements.

The official FAQ states that the total cost for regular projects is around 10,000 yuan; larger clients and complex privatization projects require a separate assessment.

How to estimate the cost of outbound calls

1. Calculate the number of customers that need to be contacted.
2. Determine the time frame that must be completed.
3. Estimate the number of calls each robot makes per day.
4. Evaluate the average connection rate and call duration.
5. Add script creation and line costs.
6. Calculate the costs for SMS messages, transfers to a human agent, and CRM.
7. Reserve costs for testing, redials, and failures.
8. Compare SaaS with pay-per-use pricing plans.
9. Ask the sales team to provide a complete written quote.

Comparison between SaaS and on-premises solutions

API and system integration

The official help center provides development resources such as AI outbound call APIs, callback functions, iFrame integration, and voice quality inspection interfaces.

• Phone numbers can be added in bulk through the interface.
• It is possible to view details of tasks and calls.
• It can receive callbacks upon the completion of a call.
• The results of outbound calls can be written into the business system.
• Pages can be embedded using iFrame.
• It can be integrated with existing live call centers.
• Authentication and throttling should be verified before enabling production access.

API Integration Tutorial

1. Apply for an open platform and a test account.
2. Retrieve authentication information from the user details.
3. Read the current API and callback documentation.
4. Create a minimal task using a test number.
5. Import of verification numbers and field mapping.
6. Check the tasks for starting and pausing operations.
7. Receive and verify the callback indicating that the call has been completed.
8. Synchronize intent tags to CRM.
9. Handle duplicate, timed-out, and failed requests.
10. Complete permission and security audits before going live.

Personal information and outbound call compliance

AI-powered outbound calls involve phone numbers, recordings, and data on customer behavior; companies must determine the legitimate source of such data, the basis for its processing, and the duration for which it should be stored.

• Do not purchase or use numbers of unknown origin.
• Marketing calls require necessary consent.
• State clearly who is calling at the start of the call.
• AI must not be used to pretend to be a regulatory body or a public institution.
• Marketing should be stopped immediately once the customer declines.
• Strictly adhere to the permitted calling hours.
• It is necessary to provide the required information prior to recording.
• Restrict employees from viewing and exporting customer data.
• Delete the information that is no longer needed once it has expired.

Requirements for preventing harassment and fraud

High concurrency capabilities are intended only for legitimate services; they must not be used for spam calls, fraudulent marketing, scams, or to bypass operators’ restrictions.

• Create business approval records for each project.
• Set a maximum limit on the number of times a single number can be contacted per day.
• Monitor the rates of complaints, rejected calls, and blocked numbers.
• Additional compliance audits are conducted for high-risk industries.
• Faking the caller ID or brand identity is prohibited.
• Any abnormal task is suspended immediately upon detection.
• Save operation logs for audit and tracing purposes.

Data security

• API keys are stored only on the server side.
• Customer numbers are masked as needed on the business interface.
• Access control is applied to recordings and transcribed text.
• Export file settings for the person in charge and expiration date.
• Employees access customer data only with the minimum level of permissions required.
• Define data boundaries before the privatization project.
• Regularly check the administrator operation logs.
• Confirm the method for deleting data after the contract ends.

APIs, privatization, and open source status

Cloud Bat Intelligence offers API services, OEM solutions, ISV partnerships, and private deployment options; however, these approaches do not mean that the products are open source.

Product advantages

• It offers comprehensive capabilities for outbound calls, inbound calls, and quality inspection.
• It supports large-model knowledge bases and multi-turn conversations.
• It provides human-machine collaboration and a mechanism for switching to manual operation.
• There are extensive call statistics and intent tags available.
• CRM and lead follow-up create a closed loop.
• Supports API, OEM, and privatization.
• The help center provides a large amount of practical information.

Usage restrictions

• Enterprise pricing should be determined by consulting demand.
• Line prices and available regions may change.
• Speech recognition is affected by accent and noise.
• Large models can still produce incorrect answers.
• Marketing outbound calls must comply with regulatory requirements.
• The capabilities on mobile devices are not as comprehensive as those on PCs.
• Private deployment does not equate to obtaining the source code.

Frequently Asked Questions

What can Yunbat Intelligence do?

It can handle tasks such as AI-driven outbound calls, intelligent inbound calls, human-machine collaboration, real-time voice customer service, call quality assessment, lead allocation, and CRM follow-up.

Can Yunbat Intelligence be tried out for free?

The official website offers a free trial and demonstration option; the specific duration of use, the number of agents available, as well as the limits on calls depend on the results of the application submitted.

How much does Yunbat Intelligent AI outbound calling cost?

According to the official FAQ, the cost of creating script texts is generally between 1,000 and 2,000 yuan; the regular call rate is around 0.1–0.15 yuan per minute. The total cost for SaaS solutions and custom projects is determined based on specific requirements.

Does Yunbat Intelligence support voice quality inspection?

It is available; it can transcribe and analyze various audio formats, and perform quality checks based on keywords, semantics, emotion, and scoring rules. The public reference price is 5 yuan per hour.

Can Yunbat Intelligence switch to a human agent?

Supported – the system can route calls to online agents based on customer preferences and business conditions; before human intervention takes place, it is possible to view the customer’s information as well as the records of previous AI-driven conversations.

Does Yunbat Intelligence offer APIs and private deployment options?

It offers interfaces such as AI-driven outbound calls, callbacks, quality inspection, and iFrame; it also supports privatization and customization. For details regarding authentication, concurrency, and pricing, please contact the official team.

Is Yunbat Intelligence an open-source product?

No, API, OEM, and private deployment do not mean that the product is open source; currently, no verifiable official source code repository for the product has been found.